Rudder

During a policy generation, we are moving around a lot of data, and some part ought to be rationalized (and even understood better)

Most notably, in prepareVariables, for each nodes and each policies, we are

• getting all variables of the policy
• checking these variables, and constructing new variables based on them

• for each of these variable, we are constructing an STVariable
• getting the escaping method for this node
• escaping the variable value based on this escaping method

We could easily get only once the escaping method for the node (this is done 6194231 times on a generation with 4500 nodes, and 1.5 millions control point), and skip the middle man and create directly the STVariable from the checked Variable

Also, policy.mergeVars looks now quite dubious